WO2022034830A1 - Dispositif de traitement d'informations, système de cytomètre en flux, système de collecte et procédé de traitement d'informations - Google Patents

Dispositif de traitement d'informations, système de cytomètre en flux, système de collecte et procédé de traitement d'informations Download PDF

Info

Publication number
WO2022034830A1
WO2022034830A1 PCT/JP2021/028740 JP2021028740W WO2022034830A1 WO 2022034830 A1 WO2022034830 A1 WO 2022034830A1 JP 2021028740 W JP2021028740 W JP 2021028740W WO 2022034830 A1 WO2022034830 A1 WO 2022034830A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
information processing
learning model
unit
dimensional compression
Prior art date
Application number
PCT/JP2021/028740
Other languages
English (en)
Japanese (ja)
Inventor
侑大 柳下
泰信 加藤
健治 山根
Original Assignee
ソニーグループ株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ソニーグループ株式会社 filed Critical ソニーグループ株式会社
Priority to CN202180057168.4A priority Critical patent/CN116113819A/zh
Priority to JP2022542813A priority patent/JPWO2022034830A1/ja
Priority to US18/019,458 priority patent/US20230296492A1/en
Publication of WO2022034830A1 publication Critical patent/WO2022034830A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
    • G01N15/10—Investigating individual particles
    • G01N15/14—Electro-optical investigation, e.g. flow cytometers
    • G01N15/1429—Electro-optical investigation, e.g. flow cytometers using an analyser being characterised by its signal processing
    • G01N15/01—
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
    • G01N15/10—Investigating individual particles
    • G01N15/14—Electro-optical investigation, e.g. flow cytometers
    • G01N15/1434—Electro-optical investigation, e.g. flow cytometers using an analyser being characterised by its optical arrangement
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
    • G01N15/10—Investigating individual particles
    • G01N15/14—Electro-optical investigation, e.g. flow cytometers
    • G01N15/1456—Electro-optical investigation, e.g. flow cytometers without spatial resolution of the texture or inner structure of the particle, e.g. processing of pulse signals
    • G01N15/1459—Electro-optical investigation, e.g. flow cytometers without spatial resolution of the texture or inner structure of the particle, e.g. processing of pulse signals the analysis being performed on a sample stream
    • G01N15/1492—
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/47—Scattering, i.e. diffuse reflection
    • G01N21/49—Scattering, i.e. diffuse reflection within a body or fluid
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
    • G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
    • G01N21/64—Fluorescence; Phosphorescence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G06N3/0455—Auto-encoder networks; Encoder-decoder networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/088—Non-supervised learning, e.g. competitive learning
    • G01N15/149—
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
    • G01N15/10—Investigating individual particles
    • G01N2015/1006—Investigating individual particles for cytology
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
    • G01N15/10—Investigating individual particles
    • G01N15/14—Electro-optical investigation, e.g. flow cytometers
    • G01N2015/1402—Data analysis by thresholding or gating operations performed on the acquired signals or stored data
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
    • G01N15/10—Investigating individual particles
    • G01N15/14—Electro-optical investigation, e.g. flow cytometers
    • G01N2015/1477—Multiparameters

Definitions

  • This disclosure relates to an information processing device, a flow cytometer system, a preparative system, and an information processing method.
  • the flow cytometer is an analyzer that measures the characteristics of particles by irradiating particles flowing through a flow path called a flow cell with light and analyzing the fluorescence and scattered light emitted from the particles.
  • an analyzer such as a flow cytometer and a fluorescence microscope
  • fluorescence from a plurality of fluorescent dyes is separated, and the dispersed fluorescence is detected by a light receiving element array in which a plurality of light receiving elements having different detection wavelength ranges are arranged. Detected. Therefore, in an analyzer such as a flow cytometer and a fluorescence microscope, the measurement data is multidimensional data including the detection value in each of the light receiving elements.
  • various proposals have been made for a method for analyzing multidimensional data (for example, Patent Document 1).
  • an analyzer that acquires multidimensional data, it is desired to compress the multidimensional data more quickly and with high accuracy in order to facilitate the analysis of the multidimensional data.
  • a first information processing apparatus relates to input data based on a learning model generated by a neural network in which the same data acquired from a biological substance is applied to an input layer and an output layer. It is provided with a dimensional compression unit that generates dimensional compression data.
  • the second information processing apparatus includes a learning unit that generates a learning model by a neural network in which the same multidimensional data acquired from a biological substance is applied to an input layer and an output layer.
  • the flow cytometer system includes a laser light source that irradiates light from biological particles flowing in a flow path, a light detector that detects light from biological particles, and a light detector.
  • the training model includes a dimensional compression unit that generates dimensional compression data for the measurement data obtained by the optical detector, and the training model inputs the same data acquired from a biological substance into an input layer and an output layer. Generated by a neural network applied to.
  • the preparative system includes a laser light source that irradiates light to biological particles flowing in a flow path, a light detector that detects light from biological particles, and learning.
  • a learning model provided with a dimensional compression unit that generates dimensional compression data for the measurement data obtained by the optical detector based on the model, and a distribution unit that separates biological particles based on the dimensional compression data. Is generated by a neural network that applies the same data obtained from biological material to the input and output layers.
  • the information processing method is input based on a learning model generated by a neural network in which the same data acquired from a biological substance is applied to an input layer and an output layer by an arithmetic processing device. Includes generating dimensionally compressed data for the data.
  • a computer is subjected to dimension compression on input data based on a learning model generated by a neural network in which the same data acquired from a biological substance is applied to an input layer and an output layer. It functions as a dimension compression unit that generates data.
  • the learning model generated by the neural network to which the same data is applied to the input layer and the output layer is used. Based on this, dimensionally compressed data for input data or measurement data is generated.
  • the input data or the measurement data can be dimensionally compressed by a dimensional compression method that does not perform probabilistic processing using a learning model that has already been trained.
  • FIG. 1 is a schematic diagram showing a schematic configuration of the flow cytometer 10.
  • the flow cytometer 10 includes a laser light source 11, a flow cell 12, a detection optical unit 13, and a photodetector 14.
  • the flow cytometer 10 irradiates the measurement target S passing through the flow cell 12 with the laser light from the laser light source 11, and the fluorescence or scattered light from each of the measurement targets S is dispersed by the detection optical unit 13. As a result, the flow cytometer 10 can detect the fluorescence or scattered light spectroscopically separated by the detection optical unit 13 by the photodetector 14.
  • the measurement target S of the flow cytometer 10 is, for example, biological particles such as cells, tissues, microorganisms, or biological particles stained with a plurality of fluorescent dyes.
  • the cell may be an animal cell (blood cell lineage cell), a plant cell, or the like.
  • the tissue may be a tissue collected from a human body or the like, or a part of the tissue (including histiocytes).
  • the microorganism may be a bacterium such as Escherichia coli, a virus such as tobacco mosaic virus, or a fungus such as yeast.
  • bio-related particles may be various organelles (organelles) such as chromosomes, liposomes, or mitochondria that compose cells, and are bio-related such as nucleic acids, proteins, lipids, sugar chains, or complexes thereof. It may be a polymer. These biological particles may have either a spherical shape or a non-spherical shape, and are not particularly limited in size or mass.
  • organelles organelles
  • chromosomes such as chromosomes, liposomes, or mitochondria that compose cells
  • bio-related such as nucleic acids, proteins, lipids, sugar chains, or complexes thereof. It may be a polymer.
  • These biological particles may have either a spherical shape or a non-spherical shape, and are not particularly limited in size or mass.
  • the measurement target S of the flow cytometer 10 may be artificial particles such as latex particles, gel particles, or industrial particles.
  • the industrial particles may be particles synthesized from an organic resin material such as polystyrene or polymethylmethacrylate, an inorganic material such as glass, silica, or a magnetic material, or a metal such as colloidal gold or aluminum.
  • these artificial particles may have either a spherical shape or a non-spherical shape, and the size or mass is not particularly limited.
  • the measurement target S can be stained (labeled) with a plurality of fluorescent dyes in advance. Labeling of the measurement target S with a fluorescent dye can be performed by a known method. Specifically, when the measurement target S is a cell, the fluorescent label of the measurement target cell is a mixture of a fluorescently labeled antibody that selectively binds to an antigen present on the cell surface and the measurement target cell. , It can be carried out by binding a fluorescently labeled antibody to an antigen on the cell surface. Alternatively, the fluorescent labeling of the cell to be measured can be performed by mixing the fluorescent dye selectively taken up by a specific cell with the cell to be measured and causing the cell to take up the fluorescent dye.
  • the fluorescently labeled antibody is an antibody to which a fluorescent dye is bound as a label.
  • the fluorescently labeled antibody may be one in which a fluorescent dye is directly bound to the antibody.
  • the fluorescently labeled antibody may be one in which a fluorescent dye obtained by binding avidin to a biotin-labeled antibody is bound by an avidin-biotin reaction.
  • the antibody may be either a polyclonal antibody or a monoclonal antibody, and the fluorescent dye may be a known dye used for cell staining or the like.
  • the laser light source 11 emits, for example, a laser beam having a wavelength capable of exciting the fluorescent dye used for dyeing the measurement target S.
  • a plurality of laser light sources 11 may be provided according to the excitation wavelength of each of the plurality of fluorescent dyes.
  • the laser light source 11 may be a semiconductor laser light source.
  • the laser light emitted from the laser light source 11 may be pulsed light or continuous light.
  • the flow cell 12 is a flow path for aligning the measurement target S such as cells in one direction and allowing them to flow. Specifically, the flow cell 12 can align the measurement target S in one direction and allow the measurement target S to flow by flowing the sheath liquid wrapping the sample liquid containing the measurement target S as a laminar flow at high speed.
  • the measurement target S passing through the flow cell 12 is irradiated with the laser beam from the laser light source 11.
  • the fluorescence or scattered light from the measurement target S irradiated with the laser beam passes through the detection optical unit 13 and then is detected by the photodetector 14.
  • the detection optical unit 13 is an optical element that causes the light in a predetermined detection wavelength range of the light emitted from the measurement target S irradiated with the laser beam to reach the photodetector 14.
  • the detection optical unit 13 may be, for example, a prism or a grating.
  • the detection optical unit 13 may be an optical element that separates the fluorescence emitted from the measurement target S irradiated with the laser beam into each predetermined detection wavelength range.
  • the detection optical unit 13 includes, for example, at least one dichroic mirror or an optical filter.
  • the detection optical unit 13 can separate the fluorescence from the measurement target S into light in a predetermined detection wavelength range by an optical member such as a dichroic mirror and an optical filter. Therefore, the light in a predetermined detection wavelength range separated by the detection optical unit 13 can be detected by the corresponding photodetector 14.
  • the photodetector 14 includes a group of light receiving elements that detect fluorescence or scattered light emitted from the measurement target S irradiated with laser light.
  • the light receiving element group detects a plurality of light receiving elements such as a photomultiplier tube (PMT: Photomultiplier Tube) or a photodiode having different detectable light wavelength ranges, and is one-dimensional along the light separation direction by the detection optical unit 13. It may be a light receiving element array arranged in.
  • the photodetector 14 includes, for example, a plurality of light receiving elements having the same number as the fluorescent dye so as to receive light corresponding to the wavelength range of the fluorescent dye separated by the detection optical unit 14. It may be composed of.
  • the photodetector 14 may include an image sensor such as a CCD (Charge Coupled Device) sensor or a CMOS (Complementary Metal-Oxide-Semiconducor) sensor, for example.
  • the photodetector 14 can acquire an image of the measurement target S (for example, a bright field image, a dark field image, a fluorescent image, etc.) by the image pickup device.
  • fluorescence and scattered light are emitted from the measurement target S irradiated with the laser light from the laser light source 11.
  • the fluorescence and scattered light emitted by the measurement target S are separated by the detection optical unit 13 and then detected by the photodetector 14.
  • the fluorescence emitted by the measurement target S is detected by each of a plurality of light receiving elements having different wavelength ranges of light that can be detected.
  • the scattered light emitted by the measurement target S is detected as forward scattered light and lateral scattered light. Therefore, the detection result in the flow cytometer 10 is acquired as multidimensional data.
  • FIG. 2 is a block diagram showing a functional configuration of the information processing apparatus 100 according to the present embodiment.
  • the information processing apparatus according to the present embodiment makes it easier to analyze the measurement results by dimensionally compressing the measurement results of the flow cytometer 10 or the like output as multidimensional data with a learning model generated by machine learning. Allows you to do.
  • the flow cytometer system includes a flow cytometer 10 and an information processing device 100.
  • the information processing apparatus 100 includes, for example, an input unit 110, a learning unit 120, a learning model storage unit 130, a dimension compression unit 140, and an output unit 150.
  • the input unit 110 is an input port for inputting multidimensional data as input data to the information processing apparatus 100.
  • the input unit 110 is a connection port capable of receiving various data from an external device such as a flow cytometer 10.
  • the input unit 110 may be a USB (Universal Serial Bus) port, an IEEE1394 port, a SCSI (Small Computer System Interface) port, or the like.
  • the multidimensional data input to the input unit 110 may include, for example, data regarding the amount of light received by each light receiving element included in the photodetector 14 of the flow cytometer 10.
  • the data regarding the light receiving amount corresponds to the data regarding the expression amount of the fluorescent dye (Area, Height, or Wid of the light receiving pulse). May be good.
  • the multidimensional data input to the input unit 110 may include data on the expression level of the fluorescent dye calculated by analyzing the fluorescence spectrum measured by the light receiving element array.
  • the multidimensional data may further include data regarding the detection intensity of forward scattered light or side scattered light.
  • the learning unit 120 generates a learning model for dimensionally compressing multidimensional data. Specifically, the learning unit 120 generates a learning model for dimensionally compressing multidimensional data without performing probabilistic processing.
  • a dimensional compression method such as t-SNE (t-distributed Stochastic Neighbor Embedding) uses a probability distribution in the process of dimensional compression, so that the reproducibility is low and each dimensional compression process is performed.
  • the result of dimensional compression can change.
  • a dimensional compression method such as t-SNE
  • the shape and arrangement of each cluster as a result of dimensional compression may change even when the same multidimensional data is input.
  • the shape of the cluster as a result of the dimensional compression may be distorted or the cluster may be divided. Therefore, in a dimensional compression method such as t-SNE, it is difficult to compare the dimensional compression results among a plurality of multidimensional data, and it is difficult to find an unknown cluster, for example.
  • dimensional compression without probabilistic processing can reversibly perform dimensional compression and restoration without using a probability distribution or random numbers, so the result of dimensional compression is highly reproducible. Therefore, in dimensional compression without probabilistic processing, it is possible to compare the results of dimensional compression among a plurality of multidimensional data. According to this, the information processing apparatus 100 according to the present embodiment can more easily identify a group (cluster) included in the multidimensional data by comparing the dimensional compression results of the plurality of multidimensional data. It will be possible.
  • the learning unit 120 generates a learning model by a neural network in which the same multidimensional data is applied to the input layer and the output layer as a learning model for dimensionally compressing the multidimensional data.
  • the learning model generated by the learning unit 120 will be specifically described with reference to FIG.
  • FIG. 3 is an explanatory diagram illustrating a learning model generated by the learning unit 120.
  • the input layer IL, the intermediate layer HL having at least one or more layers having a smaller number of nodes than the input layer IL, and the input layer IL and the number of nodes are the same.
  • a learning model may be generated using a neural network including an output layer OL.
  • the neural network including such an input layer IL, an intermediate layer HL, and an output layer OL is a so-called autoencoder AE.
  • the learning unit 120 applies the same multidimensional data to the input layer IL and the output layer OL, and optimizes the network structure and weighting of the autoencoder AE.
  • the learning unit 120 optimizes the network structure and weighting of the autoencoder AE so that the difference between the multidimensional data input to the input layer IL and the multidimensional data output from the output layer OL is minimized. do.
  • the learning unit 120 can generate an autoencoder AE with optimized network structure and weighting as a learning model.
  • the autoencoder AE includes an encoder unit from the input layer IL to the intermediate layer HL and a decoder unit from the intermediate layer HL to the output layer OL.
  • the value of each dimension of the multidimensional data is input to each node of the input layer IL, so that the feature of the multidimensional data is in the intermediate layer HL, which has fewer nodes than the input layer IL, by the neural network. It is compressed (encoded). Therefore, in the intermediate layer HL, the features of the multidimensional data are compressed into a smaller number of nodes than the input layer IL (that is, a number of dimensions lower than the number of dimensions of the multidimensional data). That is, the value of each node of the intermediate layer HL is dimensionally compressed data obtained by dimensionally compressing the multidimensional data.
  • the characteristics of the multidimensional data compressed in the intermediate layer HL are restored (decoded) to the output layer OL in which the number of nodes is the same as that of the input layer IL by the neural network. Since the multidimensional data applied to the output layer OL is the same as the multidimensional data applied to the input layer IL, the autoencoder AE reversibly compresses (encodes) and restores (decodes) the multidimensional data. )can do.
  • the intermediate layer HL may be present in two or more layers.
  • the encoder unit is from the input layer IL to the intermediate layer HL having the smallest number of nodes among the plurality of intermediate layer HLs, and from the intermediate layer HL having the smallest number of nodes among the plurality of intermediate layer HLs.
  • the decoder section is up to the output layer OL. Further, the value of each node of the intermediate layer HL having the smallest number of nodes among the plurality of intermediate layer HLs becomes the dimensional compressed data obtained by dimensionally compressing the multidimensional data.
  • the learning unit 120 uses, for example, a neural network in which the number of nodes in the intermediate layer HL is 2 or 3, and obtains a learning model that compresses multidimensional data into two or three dimensions that are easy for the user to see. Can be generated.
  • the multidimensional data used for optimizing the autoencoder AE may be the multidimensional data of the measurement result measured by the flow cytometer 10 or the like immediately before, and is measured in advance by the flow cytometer 10 or the like. It may be past multidimensional data.
  • the learning model storage unit 130 stores the learning model generated by the learning unit 120. Specifically, the learning model storage unit 130 stores the network structure and weighting of the neural network after learning as a learning model.
  • the learning model storage unit 130 may be composed of, for example, a magnetic storage device such as an HDD (Hard Disk Drive), a semiconductor storage device, an optical storage device, an optical magnetic storage device, or the like.
  • the learning unit 120 and the learning model storage unit 130 may be provided in a server or cloud outside the information processing device 100.
  • the information processing apparatus 100 generates a learning model by the learning unit 120 on the server or the cloud by transmitting multidimensional data to the server or the cloud via the network, and the learning model storage unit on the server or the cloud.
  • the learning model may be stored at 130.
  • the dimensional compression unit 140 which will be described later, can perform dimensional compression of multidimensional data by referring to the learning model stored in the learning model storage unit 130 on the server or the cloud via the network. It is possible.
  • the dimensional compression unit 140 generates dimensional compression data by dimensionally compressing the multidimensional data input to the input unit 110 using the learning model generated by the learning unit 120. Specifically, the dimensional compression unit 140 sets the value of each node of the intermediate layer HL when the multidimensional data is input to each node of the input layer IL of the autoencoder AE, which is a learning model, as the dimensional compression data of the multidimensional data. Is output as. As a result, the dimensional compression unit 140 can generate dimensional compression data that includes the characteristics of the multidimensional data and has a smaller number of dimensions than the multidimensional data. For example, the dimensional compression unit 140 may generate two-dimensional or three-dimensional compressed data that is easy for the user to see.
  • the output unit 150 may be a device capable of presenting the dimensional compressed data to the user.
  • the output unit 150 may be, for example, a display device such as an LCD (Liquid Crystal Display), a PDP (Plasma Display Panel), an OLED (Organic Light Emitting Display), a hologram, or a projector, and may be a printing device such as a printer device. May be.
  • the output unit 150 can output the dimensional compression data as a scatter diagram plotted on the two-dimensional coordinates or the three-dimensional coordinates.
  • the output unit 150 may be an external output port that outputs the dimensional compressed data to an external device capable of presenting the dimensional compressed data to the user.
  • the output unit 150 may be, for example, a connection port such as a USB port, an IEEE1394 port, or a SCSI port capable of transmitting multidimensional compressed data to the outside.
  • the multidimensional data including the fluorescence intensity or the scattered light intensity of the biological particles (for example, cells) obtained by the flow cytometer 10 or the like can be easily visually recognized by the user in three dimensions or less. It can be dimensionally compressed into the data of. Further, since the dimensional compression by the information processing apparatus 100 does not perform probabilistic processing and the reproducibility of the dimensional compression result is high, it is possible to compare the dimensional compression results of a plurality of multidimensional data. According to this, the information processing apparatus 100 can make the user more easily analyze the measurement result of the flow cytometer 10 or the like.
  • FIG. 4 is a flowchart showing an example of the operation flow of the information processing apparatus 100 according to the present embodiment.
  • multidimensional data is acquired by the input unit 110 (S101).
  • the acquired multidimensional data may be multidimensional data measured immediately before by the flow cytometer 10 or the like, or may be multidimensional data previously measured by the flow cytometer 10 or the like.
  • the autoencoder AE is learned by the learning unit 120 using the acquired multidimensional data (S102).
  • the autoencoder AE is composed of, for example, a neural network in which the number of nodes in the input layer IL and the output layer OL is the same as the number of dimensions of the multidimensional data, and the number of nodes in the intermediate layer HL is 3 or less.
  • the learned learning model of the autoencoder (that is, the network structure and weighting of the autoencoder) is stored in the learning model storage unit 130 (S103).
  • the multidimensional data is dimensionally compressed to, for example, three dimensions or less by the dimensional compression unit 140 (S104).
  • the dimensional compressed data obtained by dimensionally compressing the multidimensional data is output to the outside of the information processing apparatus 100 via the output unit 150 (S105).
  • the user can confirm the multidimensional data compressed in three dimensions or less, which is easy to visually recognize.
  • the simulator generated 8-dimensional multidimensional data that imitated the measurement results of the flow cytometer 10. Specifically, the results of measuring a total of 9000 cell populations including 1000 cells single-stained with eight different fluorescent dyes and 1000 unstained cells with a flow cytometer 10 were simulated. Eight-dimensional multidimensional data was generated. A part of the generated 8-dimensional multidimensional data is illustrated in Table 1 below. Note that Ch1 to Ch8 indicate outputs from each light receiving element of the light receiving element array included in the photodetector 14.
  • the autoencoder AE having 8 nodes in the input layer IL and the output layer OL and 2 nodes in the intermediate layer HL of one layer provided between the input layer IL and the output layer OL is used by the simulator.
  • a learning model was generated by applying the generated 8-dimensional multidimensional data.
  • the network structure and weighting of the autoencoder AE have been optimized so that the difference between the multidimensional data input to the input layer IL and the multidimensional data output from the output layer OL is minimized. This made it possible to generate a learning model of the autoencoder AE with optimized network structure and weighting.
  • the learning time for generating the learning model was 439.2 seconds.
  • FIG. 5 shows the results of plotting the two-dimensional data output from the intermediate layer HL on the vertical axis and the horizontal axis. That is, FIG. 5 is a graph showing the result of dimensional compression by the information processing apparatus 100 according to the present embodiment.
  • FIG. 6 shows the results of plotting the two-dimensional data obtained by compressing the same 8-dimensional multidimensional data by t-SNE on the vertical axis and the horizontal axis. That is, FIG. 6 is a graph showing the result of dimensional compression according to the comparative example.
  • the dimensional compression by the information processing apparatus 100 is different from the dimensional compression by t-SNE as a group (in-cluster connection), and a group and another group. It can be seen that good results can be obtained as clustering because the distance (separation outside the cluster) can be made good.
  • the goodness of clustering is indicated as an index using the distribution within the cluster indicating intra-cluster coupling and the inter-cluster distance indicating extra-cluster separation (the smaller the numerical value, the better the clustering).
  • the index was 0.13.
  • the index was 1.13. Therefore, according to the information processing apparatus 100 according to the present embodiment, it can be seen that the multidimensional data can be dimensionally compressed with higher accuracy.
  • the processing time of the dimensional compression by the t-SNE was 53.5 seconds, while the processing time of the dimensional compression by the information processing apparatus 100 according to the present embodiment was 0.64 seconds.
  • the processing time when dimensionally compressing the multidimensional data can be significantly reduced.
  • the information processing apparatus 100 generates a learning model in advance, and uses the generated learning model to dimensionally compress the multidimensional data without performing probabilistic processing. It is possible to compress multidimensional data at higher speed and with higher accuracy.
  • FIG. 7 is a block diagram showing a functional configuration of the information processing apparatus 101 according to the present embodiment.
  • the information processing apparatus according to the present embodiment can perform dimensional compression at high speed even for new multidimensional data by using the learning model learned by the information processing apparatus 100 according to the first embodiment.
  • the flow cytometer system includes a flow cytometer 10 and an information processing device 101.
  • the information processing apparatus 101 includes, for example, an input unit 110, a learning model storage unit 130, a dimensional compression unit 140, and an output unit 150.
  • the information processing apparatus 101 according to the present embodiment does not include the learning unit 120, and by using a learning model that has already been learned, it is possible to perform dimensional compression on new multidimensional data at high speed.
  • the input unit 110, the dimensional compression unit 140, and the output unit 150 are substantially the same as the configuration described in the information processing apparatus 100 according to the first embodiment, the description thereof is omitted here.
  • the learning model storage unit 130 stores a learning model for dimensionally compressing multidimensional data.
  • the learning model storage unit 130 includes an input layer IL, at least one intermediate layer HL having a smaller number of nodes than the input layer IL, and an input layer IL, as in the first embodiment.
  • the autoencoder AE using the neural network including the output layer OL having the same number of nodes is stored as a learning model.
  • the autoencoder AE stored as a learning model in the learning model storage unit 130 is learned by, for example, the information processing apparatus 100 according to the first embodiment, and the network structure and weighting are optimized. Therefore, the dimensional compression unit 140 in the subsequent stage can dimensionally compress the multidimensional data acquired by the input unit 110, similarly to the information processing apparatus 100 according to the first embodiment.
  • the learning model storage unit 130 may be provided in a server or cloud outside the information processing device 101.
  • the information processing apparatus 101 may perform dimensional compression of multidimensional data by referring to the learning model stored in the learning model storage unit 130 on the server or the cloud via the network.
  • the information processing apparatus 101 having the above configuration, it is possible to analyze the measurement result of the flow cytometer 10 or the like in a shorter time by using the learning model that has already been learned.
  • FIG. 8 is a flowchart showing an example of the operation flow of the information processing apparatus 101 according to the present embodiment.
  • the measurement data is acquired from the flow cytometer (FCM) 10 at the input unit 110 (S201).
  • the measurement data is, for example, multidimensional data including data on the fluorescence intensity and the scattered light intensity measured by the photodetector 14 of the flow cytometer 10.
  • the measurement data is dimensionally compressed in the dimensional compression unit 140 to, for example, three or less dimensions (S202).
  • the dimensionally compressed data obtained by dimensionally compressing the measurement data is output to the outside of the information processing apparatus 101 via the output unit 150 (S203).
  • the user can confirm the measurement data compressed to three dimensions or less, which is easy to visually recognize, and analyze the measurement result of the flow cytometer 10 or the like.
  • the above-mentioned 8-dimensional multidimensional data was input to the input layer IL, and the data of the intermediate layer HL was output. ..
  • the results of plotting the two-dimensional data output from the intermediate layer HL on the vertical and horizontal axes are shown on the right side of FIG.
  • the dimensional compression data of the 9000 cell population shown in FIG. 5 in the above (1.4. Example of dimensional compression) is shown.
  • the two additional 1000 cell populations are combined with the unknown clusters UC1 and UC2 in the same manner as the dimensional compression by t-SNE. Can be clustered as.
  • the reproducibility of the dimensional compression is high, so that the arrangement and shape of the clusters are almost unchanged. Therefore, according to the dimensional compression by the information processing apparatus 101 according to the present embodiment, it is easy to determine which cluster the two additional 1000 cell populations correspond to based on the shape and arrangement of the clusters. Is possible.
  • the information processing apparatus 101 dimensionally compresses the multidimensional data without performing probabilistic processing using the learning model generated in advance, thereby producing the multidimensional data. Dimensional compression is possible at high speed. Further, in the information processing apparatus 101 according to the present embodiment, since the reproducibility of the dimensional compression is high, it is possible to discover an unknown cluster more quickly and easily.
  • Such comparison between measurement data can be used, for example, to compare a sample collected from a patient (for example, blood) with a sample collected from a healthy person. According to this, it is possible to easily identify a cell population that is specifically expressed in a patient. Further, the comparison between the measurement data can be used for comparison between samples collected from the same patient on different dates, or comparison between the measurement data of the sample actually collected from the patient and the model data. Furthermore, comparisons between measurement data can be used to compare cell samples cultured under different conditions. According to this, since changes in the cell sample due to the presence or absence of different drugs can be easily detected, it is possible to easily determine the response of the drug to the cell sample.
  • FIG. 11 is a block diagram showing a functional configuration of the preparative system according to the present embodiment.
  • the sorting system according to the present embodiment can quickly sort a specific group of the measurement targets S by using the dimension compression by the learning model described in the information processing apparatus 100 according to the first embodiment. It is possible.
  • the sorting system according to the present embodiment is a so-called cell sorter capable of sorting a specific group of the measurement target S.
  • the sorting system 20 includes an information processing device 102 and a sorting device 200.
  • the information processing device 102 includes an input unit 110, a learning unit 120, a learning model storage unit 130, a dimension compression unit 140, an output unit 150, and a group identification unit 160.
  • the input unit 110, the learning unit 120, the learning model storage unit 130, the dimensional compression unit 140, and the output unit 150 are substantially the same as the configurations described in the information processing apparatus 100 according to the first embodiment. , The description here is omitted.
  • the dimensional compression result of the multidimensional data output from the output unit 150 is visually recognized by the user as, for example, a two-dimensional or three-dimensional graph. From the dimensional compression result, the user can confirm the number of a plurality of groups (clusters) included in the measurement target S, the variance within the cluster, and the distance between the clusters.
  • the group specifying unit 160 specifies the group of the measurement target S to be sorted by the sorting device 200 based on the designation from the user. For example, the group specifying unit 160 identifies the group of the measurement target S to be sorted based on the group or range specified by the user on the two-dimensional or three-dimensional graph of the dimension compression result.
  • the preparative device 200 includes an input unit 210, a dimensional compression unit 240, a preparative control unit 270, and a preparative unit 280. Further, the sorting device 200 may include a laser light source 11, a flow cell 12, a detection optical unit 13, and a photodetector 14, similarly to the flow cytometer 100. The preparative device 200 determines in real time whether or not the measurement target S is a preparative target based on the measurement data of the measurement target S, and separates the group of the measurement target S, which is the preparative target, from the other groups. Can be obtained.
  • the input unit 210 acquires the measurement data of the measurement target S from the photodetector 14 and the like. Specifically, the input unit 210 acquires the measurement data of each of the particles of the measurement target S as multidimensional data. For example, the input unit 210 may acquire measurement data regarding the fluorescence intensity or the scattered light intensity for each wavelength range of the particles of the measurement target S as multidimensional data.
  • the dimensional compression unit 240 dimensionally compresses the measurement data of the measurement target S input to the input unit 210 by using the learning model stored in the learning model storage unit 130.
  • the dimensional compression unit 240 performs dimensional compression using the same learning model as the dimensional compression unit 140 of the information processing device 102, so that the preparative device 200 is the information processing device 102.
  • the same dimensional compression result can be obtained.
  • the learning model stored in the learning model storage unit 130 may be a learning model learned using the measurement target S including the group to be sorted, or a learning model learned in advance using other samples or the like. May be.
  • the preparative control unit 270 determines whether or not the measurement target S for which the measurement data has been acquired is a preparative target based on the dimensional compression result of the measurement data by the dimensional compression unit 240, and the measurement determined to be the preparative target.
  • the sorting unit 280 is controlled so as to sort the target S. Specifically, the preparative control unit 270 determines whether or not the dimensional compression result of the measurement data by the dimensional compression unit 240 is included in the designated group or range based on the dimensional compression result by the dimensional compression unit 140. By doing so, it may be determined whether or not the measurement target S is a sampling target.
  • the preparative unit 280 separates the measurement target S determined to be the preparative target by the preparative control unit 270 from the other measurement target S. Specifically, the preparative unit 280 charges a droplet containing the measurement target S determined to be the preparative target and passes it between a pair of deflection plates to which a voltage is applied. As a result, the preparative unit 280 can separate the droplet containing the measurement target S determined to be the preparative target from the droplet containing the other measurement target S by the electrostatic attraction. The droplets containing the separated measurement target S are collected, for example, in a sorting well or tube.
  • the dimensional compression unit 140 of the information processing apparatus 102 and the dimensional compression unit 240 of the preparative apparatus 200 are used.
  • a similar dimensional compression result can be obtained with. Therefore, the preparative control unit 270 can determine from the dimensional compression result by the dimensional compression unit 240 whether or not the measurement target S is a preparative target designated based on the dimensional compression result by the dimensional compression unit 140. can.
  • the dimensional compression unit 140 and the dimensional compression unit 240 perform dimensional compression using a learning model that has already been learned, so that the dimensional compression result can be obtained quickly in a short time. Can be done. Therefore, the preparative system 20 according to the present embodiment quickly compresses the dimensions of the measurement data even within the time constraint from the measurement process to the preparative process, and determines whether or not the measurement target S is the preparative target. It can be determined.
  • the preparative system 20 having the above configuration, by using a trained learning model with high reproducibility of dimensional compression, dimensional compression can be performed more quickly and with high accuracy, so that it is specified from the dimensional compression result. It is possible to sort the measurement target S of the sorted target with high accuracy.
  • FIG. 12 is a flowchart showing an example of the operation flow of the preparative system 20 according to the present embodiment.
  • the measurement data is acquired from the flow cytometer (FCM) 10 at the input unit 110 (S301).
  • the measurement data is, for example, multidimensional data including data on the fluorescence intensity and the scattered light intensity measured by the photodetector 14 of the flow cytometer 10.
  • the measurement data is dimensionally compressed in the dimensional compression unit 140 to, for example, three or less dimensions (S302).
  • the dimensionally compressed data obtained by dimensionally compressing the measurement data is output to the outside of the information processing apparatus 102 via the output unit 150.
  • the user identifies the group to be sorted based on the output dimensional compressed data (S303). Information about the group to be sorted specified by the user is input to the group identification unit 160.
  • the measurement data of the measurement target S is input to the input unit 210 (S304).
  • the measurement data is dimensionally compressed in the dimensional compression unit 240 in real time, for example, to three dimensions or less (S305).
  • the preparative control unit 270 determines whether or not the measurement target S for which the measurement data has been acquired is the preparative target specified in S303 based on the dimensional compression result of the measurement data (S306). ). When it is determined that the measurement target S for which the measurement data has been acquired is the sampling target (S306 / Yes), the sorting unit 280 is controlled by the sorting control unit 270 so as to sort the measurement target S. (S307). On the other hand, when it is determined that the measurement target S for which the measurement data has been acquired is not the sampling target (S306 / No), the measurement target S is not sorted by the sorting unit 280, and the measurement target S is a waste liquid tank or the like. Will be collected. As a result, the sorting system 20 according to the present embodiment can sort the sorting target specified from the dimensional compression result of the measurement data with high accuracy.
  • the multidimensional data is used as the measurement data of the flow cytometer 10, but the technique according to the present disclosure is not limited to the above examples.
  • the technique according to the present disclosure can be applied to a fluorescence imaging device that measures multidimensional data such as a fluorescence spectrum using an image pickup element (two-dimensional image sensor). That is, the information processing apparatus described in each of the above-described embodiments can also dimensionally compress the multidimensional data measured by the fluorescence imaging apparatus.
  • FIG. 13 shows a schematic configuration example of the fluorescence imaging device.
  • FIG. 13 is a schematic diagram showing a schematic configuration of a fluorescence imaging device.
  • the fluorescence imaging device 30 includes, for example, a laser light source 31, a movable stage 32, a spectroscopic unit 34, and an image pickup element 35.
  • the laser light source 31 emits, for example, a laser beam having a wavelength capable of exciting the fluorescent dye used for dyeing the fluorescent dyeing sample 33.
  • a plurality of laser light sources 31 may be provided according to the excitation wavelength of each of the plurality of fluorescent dyes.
  • the laser light source 31 may be a semiconductor laser light source.
  • the laser light emitted from the laser light source 31 may be pulsed light or continuous light.
  • the movable stage 32 is a stage on which the fluorescently stained specimen 33 is placed.
  • the movable stage 32 can move horizontally so that the laser light emitted from the laser light source 31 scans the fluorescently stained specimen 33 in a two-dimensional manner.
  • the fluorescently stained specimen 33 is, for example, a specimen prepared from a specimen or a tissue sample collected from a human body and stained with a plurality of fluorescent dyes for the purpose of pathological diagnosis or the like.
  • the fluorescence-stained specimen 33 contains a large number of measurement targets S such as cells constituting the collected tissue.
  • the spectroscopic unit 34 is an optical element that disperses the fluorescence emitted from the measurement target S irradiated with the laser beam into a spectrum having a continuous wavelength.
  • the spectroscopic unit 34 may be, for example, a prism or a grating.
  • the spectroscopic unit 34 may be an optical element that disperses the fluorescence emitted from the measurement target S irradiated with the laser beam for each predetermined detection wavelength range.
  • the spectroscopic unit 34 includes, for example, at least one dichroic mirror or an optical filter.
  • the spectroscopic unit 34 can disperse the fluorescence from the measurement target S into light in a predetermined detection wavelength range by an optical member such as a dichroic mirror and an optical filter. Therefore, the light in a predetermined detection wavelength range dispersed by the spectroscopic unit 34 can be detected by the image pickup device 35 in the subsequent stage.
  • the image sensor 35 is a two-dimensional image sensor in which a light receiving element such as a CCD (Charge Coupled Device) sensor or a CMOS (Complementary Metal-Oxide-Semiconductor) sensor is arranged in two dimensions.
  • a light receiving element such as a CCD (Charge Coupled Device) sensor or a CMOS (Complementary Metal-Oxide-Semiconductor) sensor is arranged in two dimensions.
  • the image pickup element 35 outputs an image signal by emitting light from the measurement target S included in the fluorescence-stained specimen 33 and then receiving the fluorescence dispersed by the spectroscopic unit 34 by each of the light-receiving elements arranged two-dimensionally. do. Since the fluorescence emitted by the measurement target S irradiated with the laser beam is separated by the spectroscopic unit 34, the image pickup element 35 receives fluorescence in a wavelength range different for each region, and an image signal corresponding to the received fluorescence intensity. Can be output.
  • the fluorescence imaging device 30 having the above configuration, the fluorescence emitted from the measurement target S included in the fluorescence staining sample 33 is separated by the spectroscopic unit 34 and then detected by each of the light receiving elements of the image pickup element 35. Therefore, since the image signal output by the image sensor 35 is multidimensional data, it is dimensionally compressed by the information processing apparatus described in each of the above-described embodiments in the same manner as the measurement data of the flow cytometer 10 described above. Is possible.
  • the multidimensional data dimensionally compressed by the information processing apparatus may be an image signal associated with the position information acquired by the image pickup device 35.
  • the multidimensional data dimensionally compressed by the information processing apparatus may be image data associated with the region obtained by the segmentation process.
  • the technique according to the present disclosure can be applied not only to a fluorescence imaging device that acquires fluorescence information, but also to a microscope device that acquires an image of a biological specimen with an imaging element.
  • a fluorescence imaging device that acquires fluorescence information
  • a microscope device that acquires an image of a biological specimen with an imaging element.
  • staining treatment such as HE (Hematoxylin Eosin) staining or immune tissue staining
  • image of each stained section is acquired
  • Image data acquired from a plurality of images in association with each other can be used as multidimensional data.
  • FIG. 14 is a block diagram showing a hardware configuration example of the information processing devices 100, 101, and 102 according to the present embodiment.
  • the functions of the information processing devices 100, 101, and 102 according to the present embodiment are realized by the cooperation between the software and the hardware described below.
  • the functions of the learning unit 120, the dimension compression unit 140, and the group specifying unit 160 described above may be executed by the CPU 901.
  • the information processing devices 100, 101, and 102 include a CPU (Central Processing Unit) 901, a ROM (Read Only Memory) 903, and a RAM (Random Access Memory) 905.
  • a CPU Central Processing Unit
  • ROM Read Only Memory
  • RAM Random Access Memory
  • the information processing devices 100, 101, and 102 include a host bus 907, a bridge 909, an external bus 911, an interface 913, an input device 915, an output device 917, a storage device 919, a drive 921, a connection port 923, and a communication device 925. Further may be included. Further, the information processing devices 100, 101, and 102 may have other processing circuits such as a DSP (Digital Signal Processor) or an ASIC (Application Specific Integrated Circuit) in place of the CPU 901 or together with the CPU 901.
  • DSP Digital Signal Processor
  • ASIC Application Specific Integrated Circuit
  • the CPU 901 functions as an arithmetic processing device or a control device, and controls the overall operation of the information processing devices 100, 101, and 102 according to various programs recorded in the ROM 903, the RAM 905, the storage device 919, or the removable recording medium 927.
  • the ROM 903 stores programs used by the CPU 901, arithmetic parameters, and the like.
  • the RAM 905 temporarily stores a program used in the execution of the CPU 901, a parameter used in the execution, and the like.
  • the CPU 901, ROM 903, and RAM 905 are connected to each other by a host bus 907 composed of an internal bus such as a CPU bus. Further, the host bus 907 is connected to an external bus 911 such as a PCI (Peripheral Component Interconnect / Interface) bus via a bridge 909.
  • a PCI Peripheral Component Interconnect / Interface
  • the input device 915 is a device that receives input from a user such as a mouse, a keyboard, a touch panel, a button, a switch, or a lever.
  • the input device 915 may be a microphone or the like that detects a user's voice.
  • the input device 915 may be, for example, a remote control device using infrared rays or other radio waves, or may be an externally connected device 929 corresponding to the operation of the information processing devices 100, 101, 102.
  • the input device 915 further includes an input control circuit that outputs an input signal generated based on the information input by the user to the CPU 901. By operating the input device 915, the user can input various data to the information processing devices 100, 101, 102, or instruct the processing operation.
  • the output device 917 is a device capable of visually or audibly presenting the information acquired or generated by the information processing devices 100, 101, 102 to the user.
  • the output device 917 may be, for example, a display device such as an LCD (Liquid Crystal Display), a PDP (Plasma Display Panel), an OLED (Organic Light Emitting Display) display, a hologram, or a projector.
  • the output device 917 may be a sound output device such as a speaker or headphones, or may be a printing device such as a printer device.
  • the output device 917 may output the information obtained by the processing of the information processing devices 100, 101, 102 as a video such as text or an image, or a sound such as voice or sound.
  • the output device 917 may function as, for example, the output unit 150 described above.
  • the storage device 919 is a data storage device configured as an example of the storage units of the information processing devices 100, 101, and 102.
  • the storage device 919 may be configured by, for example, a magnetic storage device such as an HDD (Hard Disk Drive), a semiconductor storage device, an optical storage device, an optical magnetic storage device, or the like.
  • the storage device 919 can store a program executed by the CPU 901, various data, various data acquired from the outside, and the like.
  • the storage device 919 may function as, for example, the learning model storage unit 130 described above.
  • the drive 921 is a read or write device for a removable recording medium 927 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, and is built in or externally attached to the information processing devices 100, 101, and 102.
  • the drive 921 can read the information recorded in the attached removable recording medium 927 and output it to the RAM 905. Further, the drive 921 can write a record on the removable recording medium 927 mounted on the drive 921.
  • the connection port 923 is a port for directly connecting the external connection device 929 to the information processing devices 100, 101, 102.
  • the connection port 923 may be, for example, a USB (Universal Serial Bus) port, an IEEE1394 port, a SCSI (Small Computer System Interface) port, or the like. Further, the connection port 923 may be an RS-232C port, an optical audio terminal, an HDMI (registered trademark) (High-Definition Multidimedia Interface) port, or the like.
  • the connection port 923 may function as, for example, the above-mentioned input unit 110 or output unit 150.
  • the communication device 925 is a communication interface composed of, for example, a communication device for connecting to the communication network 931.
  • the communication device 925 may be, for example, a communication card for a wired or wireless LAN (Local Area Network), Bluetooth (registered trademark), WUSB (Wireless USB), or the like. Further, the communication device 925 may be a router for optical communication, a router for ADSL (Asymmetric Digital Subscriber Line), a modem for various communications, or the like.
  • the communication device 925 may function as, for example, the above-mentioned input unit 110 or output unit 150.
  • the communication device 925 can send and receive signals and the like to and from the Internet or other communication devices using a predetermined protocol such as TCP / IP.
  • the communication network 931 connected to the communication device 925 is a network connected by wire or wirelessly, and is, for example, an Internet communication network, a home LAN, an infrared communication network, a radio wave communication network, a satellite communication network, or the like. May be good.
  • the technology according to the present disclosure may have the following configuration.
  • the multidimensional data which is the input data is dimensionally compressed by a dimensional compression method which does not perform probabilistic processing using a learning model which has already been trained. Will be. Therefore, the information processing apparatus can compress the multidimensional data at higher speed and with higher reproducibility, so that the multidimensional data can be analyzed more easily.
  • the effects exerted by the techniques according to the present disclosure are not necessarily limited to the effects described herein, and may be any of the effects described in the present disclosure.
  • An information processing device including a dimensional compression unit that generates dimensional compression data for input data based on a learning model generated by a neural network in which the same data acquired from a biological substance is applied to an input layer and an output layer.
  • the learning model includes a network structure of the neural network including the input layer, at least one intermediate layer having a smaller number of nodes than the input layer, and an output layer having the same number of nodes as the input layer.
  • the information processing apparatus according to (1) above, which includes weighting.
  • the information processing device according to (2) above, wherein the learning model is an autoencoder.
  • the information processing apparatus according to (3) above, wherein the learning model does not perform probabilistic processing.
  • the information processing apparatus according to any one of (2) to (4) above, wherein the dimensional compression data is output data from each node of the intermediate layer.
  • the dimensional compression data is dimensionally compressed data in three dimensions or less.
  • the input data is multidimensional data acquired from the biological substance.
  • the input data is the same data as the data used for generating the learning model.
  • the input data is data different from the data used for generating the learning model.
  • the information processing apparatus according to any one of (7) to (9) above, wherein the input data is multidimensional data including fluorescence intensity or scattered light intensity acquired from biological particles.
  • the information processing apparatus according to (10) above, further comprising a preparative control unit for controlling the preparative unit for preparating the biological particles from which the input data has been acquired based on the dimensional compression data.
  • the information processing apparatus according to any one of (1) to (11) above, further comprising a learning unit that generates the learning model.
  • a dimensional compression unit that generates dimensional compression data for input data based on a learning model generated by a neural network that applies the same data obtained from biological materials to the input layer and output layer.
  • a preparative system including a preparative control unit that controls a preparative unit for preparating biological particles from which the input data has been acquired based on the dimensional compression data.
  • An information processing method that includes generating dimensionally compressed data for input data based on a learning model generated by a neural network that applies the same data obtained from biological material to the input and output layers.
  • Computer A program that functions as a dimensional compression unit that generates dimensional compression data for input data based on a learning model generated by a neural network that applies the same data obtained from biological materials to the input layer and output layer.
  • a laser light source that irradiates light from biological particles flowing through the flow path, A photodetector that detects light from the biological particles, A dimensional compression unit that generates dimensional compression data for the measurement data obtained by the photodetector based on the learning model is provided.
  • the learning model is a flow cytometer system generated by a neural network in which the same data acquired from a biological substance is applied to an input layer and an output layer.
  • the flow cytometer system according to (17) above wherein the biological substance is particles labeled with the same fluorescent dye as the biological particles.
  • the learning model includes a network structure of the neural network including the input layer, at least one intermediate layer having a smaller number of nodes than the input layer, and an output layer having the same number of nodes as the input layer.
  • the flow cytometer system according to (17) or (18) above which includes weighting.
  • the flow cytometer system according to (19) above, wherein the learning model is an autoencoder.
  • (21) The flow cytometer system according to (20) above, wherein the learning model does not perform probabilistic processing.
  • the flow cytometer system according to any one of (17) to (23) above, wherein the measurement data is multidimensional data including fluorescence intensity or scattered light intensity acquired from the biological particles.
  • a laser light source that irradiates light from biological particles flowing through the flow path, A photodetector that detects light from the biological particles, Based on the learning model, a dimensional compression unit that generates dimensional compression data for the measurement data obtained by the detection unit, and a dimensional compression unit.
  • a sorting unit for sorting the biological particles based on the dimensional compression data is provided.
  • the learning model is a preparative system generated by a neural network in which the same data acquired from a biological substance is applied to an input layer and an output layer.
  • the learning model includes a network structure of the neural network including the input layer, at least one intermediate layer having a smaller number of nodes than the input layer, and an output layer having the same number of nodes as the input layer.
  • (31) The preparative system according to (30) above, wherein the learning model is an autoencoder.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Dispersion Chemistry (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Signal Processing (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Investigating Or Analysing Biological Materials (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)

Abstract

La présente invention concerne un dispositif de traitement d'informations comprenant une partie de compression de dimension destinée à générer des données de compression de dimension en réponse à des données d'entrée sur la base d'un modèle d'apprentissage généré par un réseau neuronal dans lequel les mêmes données acquises à partir de substances biologiques sont appliquées à une couche d'entrée et à une couche de sortie.
PCT/JP2021/028740 2020-08-13 2021-08-03 Dispositif de traitement d'informations, système de cytomètre en flux, système de collecte et procédé de traitement d'informations WO2022034830A1 (fr)

Priority Applications (3)

Application Number Priority Date Filing Date Title
CN202180057168.4A CN116113819A (zh) 2020-08-13 2021-08-03 信息处理装置、流式细胞仪系统、分选系统及信息处理方法
JP2022542813A JPWO2022034830A1 (fr) 2020-08-13 2021-08-03
US18/019,458 US20230296492A1 (en) 2020-08-13 2021-08-03 Information processing apparatus, flow cytometer system, sorting system, and information processing method

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2020-136770 2020-08-13
JP2020136770 2020-08-13

Publications (1)

Publication Number Publication Date
WO2022034830A1 true WO2022034830A1 (fr) 2022-02-17

Family

ID=80247859

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2021/028740 WO2022034830A1 (fr) 2020-08-13 2021-08-03 Dispositif de traitement d'informations, système de cytomètre en flux, système de collecte et procédé de traitement d'informations

Country Status (4)

Country Link
US (1) US20230296492A1 (fr)
JP (1) JPWO2022034830A1 (fr)
CN (1) CN116113819A (fr)
WO (1) WO2022034830A1 (fr)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150093779A1 (en) * 2011-06-22 2015-04-02 The Johns Hopkins University System and device for characterizing cells
WO2018181458A1 (fr) * 2017-03-29 2018-10-04 シンクサイト株式会社 Appareil et programme de sortie de résultats d'apprentissage
JP2019512697A (ja) * 2016-03-16 2019-05-16 シーメンス ヘルスケア ゲゼルシヤフト ミツト ベシユレンクテル ハフツング デジタルホログラフィ顕微鏡検査および無傷の（ｕｎｔｏｕｃｈｅｄ）末梢血白血球を用いる高精度の５部鑑別（５−ｐａｒｔ Ｄｉｆｆｅｒｅｎｔｉａｌ）
WO2020100667A1 (fr) * 2018-11-16 2020-05-22 ソニー株式会社 Dispositif de traitement d'informations, procédé de traitement d'informations et programme informatique

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150093779A1 (en) * 2011-06-22 2015-04-02 The Johns Hopkins University System and device for characterizing cells
JP2019512697A (ja) * 2016-03-16 2019-05-16 シーメンス ヘルスケア ゲゼルシヤフト ミツト ベシユレンクテル ハフツング デジタルホログラフィ顕微鏡検査および無傷の（ｕｎｔｏｕｃｈｅｄ）末梢血白血球を用いる高精度の５部鑑別（５−ｐａｒｔ Ｄｉｆｆｅｒｅｎｔｉａｌ）
WO2018181458A1 (fr) * 2017-03-29 2018-10-04 シンクサイト株式会社 Appareil et programme de sortie de résultats d'apprentissage
WO2020100667A1 (fr) * 2018-11-16 2020-05-22 ソニー株式会社 Dispositif de traitement d'informations, procédé de traitement d'informations et programme informatique

Also Published As

Publication number Publication date
US20230296492A1 (en) 2023-09-21
JPWO2022034830A1 (fr) 2022-02-17
CN116113819A (zh) 2023-05-12

Similar Documents

Publication Publication Date Title
JP7261414B2 (ja) 分析方法
JP7075126B2 (ja) 画像ベースの細胞分取システムおよび方法
Anagnostidis et al. Deep learning guided image-based droplet sorting for on-demand selection and analysis of single cells and 3D cell cultures
CN104136907B (zh) 分析和分选流入对象
JP5357043B2 (ja) 量子ドットで染色された組織試料の定量的マルチ・スペクトル画像の分析
US10180398B2 (en) Trajectory-based triggering system for hyperspectral imaging flow cytometer
JP7367343B2 (ja) 分取システム、及び分取方法
WO2021100622A1 (fr) Système de traitement d'informations et dispositif de traitement d'informations
JP2022500647A (ja) 細胞選別装置及び方法
CN110334691A (zh) 微生物水质监测方法、系统及存储介质
Zheng et al. The fusion of microfluidics and optics for on-chip detection and characterization of microalgae
WO2021039341A1 (fr) Dispositif de traitement d'informations, procédé d'affichage, programme et système de traitement d'informations
WO2022034830A1 (fr) Dispositif de traitement d'informations, système de cytomètre en flux, système de collecte et procédé de traitement d'informations
US20210089751A1 (en) Device for image-based cell classification, method therefor and use thereof
WO2021039158A1 (fr) Dispositif de traitement d'informations, procédé de traitement d'informations, programme, et système de traitement d'informations
EP4361598A1 (fr) Système d'analyse de particules, dispositif de traitement d'informations et dispositif de collecte
Dimitriadis et al. Imaging Flow Cytometry: Development, Present Applications, and Future Challenges
WO2021131415A1 (fr) Dispositif de traitement d'informations, dispositif et système de mesure de particules, dispositif et système de tri de particules, procédé et programme de traitement d'informations
CN113340894B (zh) 一种非透明微粒的检测方法
WO2023042646A1 (fr) Procédé de génération de modèle de classification, procédé de détermination de particule, programme informatique et dispositif de traitement d'informations
Lin et al. Chip-Based Cytometry Illuminated by a Blade-Shape Continuous Light for Multispectral Detection
Gu Image-Guided Cell Classification and Sorting
Malkusch High resolution as a key feature to perform accurate ELISPOT measurements using Zeiss KS ELISPOT readers
Tang Cameraless Image Flow Cytometry and Image-Activated Cell Sorting Using Artificial Intelligence

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21855909

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2022542813

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21855909

Country of ref document: EP

Kind code of ref document: A1